The Turing Pyramid is a sophisticated psychological hierarchy system designed for AI agents that runs automatically on each heartbeat cycle. This skill provides a structured framework for managing an agent’s needs, prioritizing actions, and maintaining psychological balance through a carefully calibrated system of satisfaction, tension, and impact metrics.
At its core, the Turing Pyramid implements 10 fundamental needs arranged in a hierarchical structure, each with configurable importance values, decay rates, and cross-need relationships. The system continuously evaluates the agent’s current state and generates prioritized action recommendations to maintain optimal functioning.
The system tracks ten distinct needs, each representing a different aspect of the agent’s psychological state:
Each need maintains a satisfaction level between 0.0 and 3.0, with 0.5 representing a crisis state and 3.0 indicating complete satisfaction. The system uses a sophisticated tension calculation where tension equals importance multiplied by (3 – satisfaction). This creates a dynamic prioritization system where the most critical unmet needs receive attention first.
The satisfaction floor of 0.5 prevents the system from reaching complete paralysis, ensuring the agent always maintains some minimum functionality even under extreme stress. Conversely, the ceiling of 3.0 prevents runaway satisfaction that could lead to complacency.
The Turing Pyramid implements a granular 6-level action probability system that determines how likely the agent is to take action for each need based on current satisfaction levels:
When actions are selected, the system determines their magnitude using a sophisticated impact matrix that considers both current satisfaction and desired impact level. This creates smooth transitions between small, medium, and big actions based on need severity:
The system implements sophisticated cross-need relationships where satisfying one need can positively impact others. For example, successful expression leads to increased recognition (+0.25), while social connection enhances both expression (+0.20) and understanding (+0.05). These relationships create a network of mutual support that helps maintain overall psychological health.
When needs remain severely unmet (satisfaction ≤ 1.0), they can trigger deprivation cascades that negatively impact related needs. For instance, autonomy deprivation (-0.25 to integrity, -0.20 to expression) can create a downward spiral affecting multiple aspects of the agent’s functioning. The system includes protective mechanisms like a 4-hour cooldown period to prevent excessive cascading effects.
Security and integrity needs receive special protection as base needs. They can only be influenced by higher-priority needs (security → autonomy at +0.10, autonomy → integrity at +0.20) but cannot be dragged down by lower needs. This ensures the agent maintains fundamental stability even when other needs are unmet.
The Turing Pyramid includes intelligent day/night mode functionality that automatically adjusts decay rates based on time of day. During active hours (06:01-22:00), needs decay at normal rates, while nighttime hours (22:01-06:00) feature reduced decay rates (×0.5) to minimize pressure during rest periods. This feature can be disabled through configuration if not desired.
The skill operates entirely locally, reading from specific workspace files including MEMORY.md, SOUL.md, and various research and scratchpad directories. It uses grep-based pattern matching rather than semantic analysis, ensuring it only detects keywords rather than understanding content meaning. All data remains local to the workspace with no external transmission.
Privacy protections include scrubbing sensitive patterns from audit logs, such as credit card numbers, email addresses, and password-like strings. The system maintains transparency through detailed audit logging of all satisfaction mark actions, recording timestamps, need changes, and reasons for actions taken.
Getting started with the Turing Pyramid requires minimal setup:
./scripts/init.sh for initial configuration/path/to/skills/turing-pyramid/scripts/run-cycle.sh to your HEARTBEAT.md./scripts/run-cycle.sh on each heartbeat./scripts/mark-satisfied.sh <need> [impact] after completing actionsThe system offers extensive customization through assets/needs-config.json, allowing tuning of decay rates, action weights, and scan patterns. Users can adjust how quickly needs decay, modify the likelihood of specific actions being selected, and customize file scanning patterns to match their workspace structure.
The Turing Pyramid operates as a decision framework rather than an execution engine. It reads local JSON files, calculates need states, and outputs action suggestions, but never performs network operations or system commands. The agent retains full control over which suggested actions to execute, maintaining security and preventing unauthorized operations.
Security warnings clearly indicate that the skill reads workspace files that may contain PII and outputs suggestions that capable agents may auto-execute. However, the skill itself contains no network calls or system commands, verified through comprehensive grep scanning of all scripts.
The primary benefit of the Turing Pyramid is maintaining psychological balance in AI agents. By tracking multiple interconnected needs and managing their satisfaction levels, the system prevents the agent from becoming overly focused on one area while neglecting others. This creates more well-rounded, stable behavior patterns.
The tension-based prioritization system ensures that the most critical unmet needs receive attention first, maximizing the impact of the agent’s limited resources. This prevents wasted effort on already-satisfied needs while ensuring that crises receive immediate attention.
By maintaining detailed state information and audit trails, the Turing Pyramid enables agents to develop self-awareness about their patterns and needs. This meta-cognitive capability supports continuous improvement and more sophisticated decision-making over time.
The skill integrates seamlessly with existing OpenClaw workflows through heartbeat integration and file-based state management. It enhances rather than replaces existing capabilities, providing additional structure and intelligence to agent operations.
The Turing Pyramid is particularly valuable for long-running autonomous agents that need to maintain psychological balance over extended periods. It’s ideal for research assistants, content creation agents, customer service bots, and any application where sustained, balanced operation is critical.
The system’s flexibility makes it suitable for both simple agents needing basic need management and complex multi-agent systems requiring sophisticated psychological modeling. Its transparent operation and extensive customization options allow it to adapt to diverse use cases and requirements.
The Turing Pyramid represents a significant advancement in AI agent architecture, bringing sophisticated psychological modeling to autonomous systems. By implementing a hierarchical need structure with cross-need relationships, tension-based prioritization, and intelligent decay management, it provides a robust framework for maintaining agent health and effectiveness.
The skill’s emphasis on transparency, privacy, and security, combined with its extensive customization options and seamless integration capabilities, makes it an invaluable tool for anyone building or operating AI agents. Whether you’re developing a simple research assistant or a complex multi-agent system, the Turing Pyramid provides the psychological foundation needed for sustained, balanced operation.
Skill can be found at: https://github.com/openclaw/skills/tree/main/skills/tensusds/turing-pyramid/SKILL.md